OpenAI Says Its New Model Might Be AGI. Wall Street Is Already Asking Who Gets Hurt.

OpenAI released a new flagship model this week and, in the company's own framing, described it as an early version of artificial general intelligence — a term that until recently was reserved for theoretical debates, not product launches. Whatever you make of that claim, the market reaction was immediate and specific: shares of two companies that make chip-design software fell within hours of a demo video showing the new model laying out a printed circuit board on its own, a task that has historically required skilled human engineers and represents a meaningful bottleneck in electronics manufacturing.

This is worth sitting with for a moment, because it's a cleaner example than usual of how AI progress is starting to hit the real economy in narrow, unpredictable places rather than broad, predictable ones. Nobody was running a hedge on electronic design automation stocks a week ago. Then a 15-second video showed software doing autonomous circuit board layout, and two established companies in that niche took a hit the same morning, even as the broader semiconductor sector rallied. The lesson isn't that this particular software category is doomed. It's that the next casualty of AI automation is unlikely to announce itself in advance.

The skepticism industry moved fast, too. A prominent Wall Street research note flagged the same week that AI token pricing — essentially the cost of running these models at scale — is collapsing amid frontier competition, with one index of data pricing down nearly 30% in a single month. That's a genuine tension at the center of the AI trade: the technology keeps getting more capable, which should in theory support higher spending, but the unit economics of actually running it keep getting cheaper, which squeezes the margins of anyone selling access to it. More demand and falling prices don't automatically net out to more revenue, and that arithmetic is exactly what's kept a running debate alive over how much of current AI infrastructure spending eventually earns its keep.

Meanwhile the capital story keeps getting bigger regardless of that debate. One of the leading AI labs is reportedly on a path to grow its computing capacity roughly fifteen-fold by the end of the decade, backed by a cascade of compute, chip, and power agreements that increasingly resemble utility-scale infrastructure deals rather than software contracts. That same lab's initial public offering is expected within weeks and could set a fresh record for the largest public listing ever, following a similar record set earlier this year by another company in the space. When a private company's IPO prospectus becomes one of the most closely watched financial documents of the year, that itself tells you how central this sector has become to the broader market's health — a health that, judging by recent Treasury market strain, is increasingly being financed with debt.

For everyday consumers and workers, the through-line is less about which chatbot is smartest and more about which job categories quietly become optional. Automated circuit board layout is a small, technical example, but it's the same mechanism that's already reshaping customer service, coding, and now design engineering — narrow tasks peeled off one at a time, each individually unremarkable, cumulatively significant. The companies selling the infrastructure for this transition are richly rewarded in the meantime; the workers whose tasks get automated rarely see it coming until the software demo goes viral.

None of this means the technology isn't real or that its long-run economic benefits won't be substantial — plenty of credible voices argue AI's ultimate value to society will exceed what any single company captures in market value. But real and profitable are different claims, and 2026's version of the AI trade increasingly asks investors, workers, and voters to hold both possibilities at once: transformative technology, and a capital structure built on debt and margin compression that hasn't yet been tested by a downturn.